Collection of scripts to reproduce experiments presented in the paper.
Run all scripts from project root directory (containing License, setup.py etc), i.e. /path/to/rl_cbf
Run experiments on CartPole:
bash experiments/training/cartpole_multiple_seeds.sh Here we explain how to reproduce figures from the paper.
One pre-trained checkpoint per experimental setting is available in experiments/artifacts/checkpoints.
Metrics for 5 seeded runs of each experimental setting is available in experiments/artifacts/data/metrics.csv
Collection of pre-generated plots is available in experiments/artifacts/plots. They are generated as follows:
bash experiments/generate_plots.sh